• Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
No Downloads

Views

Total Views
3,423
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
76
Comments
0
Likes
6

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Freebase Schema Jamie TaylorWednesday, December 8, 2010
  • 2. Goals • Schema: The Freebase Data Model • Schema as API • Schema patternsWednesday, December 8, 2010
  • 3. Freebase is a collection of facts Sofia Coppola directed Marie Antoinette { { { Freebase only contains nodes and LinksWednesday, December 8, 2010
  • 4. Freebase is a GraphWednesday, December 8, 2010
  • 5. Freebase is a labeled Graph child wrote directed parent sibling directed starred_in directed starred_inWednesday, December 8, 2010
  • 6. Schema "All the things you can say about something in Freebase" Schema is the data model for FreebaseWednesday, December 8, 2010
  • 7. All nodes are “/type/object” /m/02vyw /people/person id type type /film/director [{ /type/object/name   "id":"/m/02vyw", name   "name":null,   "type":[{}] “Francis Coppola” }]Wednesday, December 8, 2010
  • 8. Types suggest properties to use /m/02vyw id type /film/director lm /fi or ct ire /d lm /fi /type/object/id /en/bram_stokers_draculaWednesday, December 8, 2010
  • 9. Queries follow schema [{    "id": "/en/francis_ford_coppola",    "/film/director/film": [{      "id":null, "name":null    }] }]Wednesday, December 8, 2010
  • 10. Properties link the graph together /m/02vyw id type /film/director lm /fi or ct ire ten _by /d wri t lm /fi /type/object/id /en/bram_stokers_draculaWednesday, December 8, 2010
  • 11. Queries follow schema [{    "id": "/en/francis_ford_coppola", Name is returned (how to get ID?)    "/film/director/film": [{ How to get all the      "id": "/en/bram_stokers_dracula", writters for all of Coppola’s movies? "written_by":null    }] }]Wednesday, December 8, 2010
  • 12. Core ConceptsWednesday, December 8, 2010
  • 13. Core Concepts Instance: • Topic: "a thing in the world" • Blade Runner, Ridley Scott, NBC, Last Proof Schema: • Types - Categorical collections of instances • Properties - Relationships between instancesWednesday, December 8, 2010
  • 14. Core Concepts An instance may have multiple Types • "Co-Types" (Types are mix-ins) • Arnold Schwartzeneger • Person, Actor, Politician, Sports FigureWednesday, December 8, 2010
  • 15. Lessons from everyday vocabulary Wikipedia Word Frequency 20000000 18000000 16000000 14000000 12000000 Frequency 10000000 8000000 6000000 4000000 2000000 0 0 20 40 60 80 100 120 Rank Data from Victor S. GrishchenkoWednesday, December 8, 2010
  • 16. Schema Principle #1 Event Example: -Film Festival -Battle of Waterloo Use Co-Types Liberally: Use a few large, encompassing Types to provide general information Use several smaller, fine grained Types to provide detailed informationWednesday, December 8, 2010
  • 17. Core Concepts Relational DB vs RDF Properties are defined on Types • Properties are the vocabulary for a specific Type • An instance must be “an instance of a type” before it can use the Type’s properties to describe itselfWednesday, December 8, 2010
  • 18. Core Concepts Expected Type ~ RDFS Range • A Property Value has a specific Type • "Expected Type" • A Property has exactly one Expected Type ManufacturesWednesday, December 8, 2010
  • 19. Core Concepts Expected Types (Property Values): • Value Types (literals) • String (two flavors), Integer, Float, DateTime, boolean • Object Types • Everything ElseWednesday, December 8, 2010
  • 20. /type/object contrast to common topic Everything in Freebase has this Type Provides basic properties • Type • Name • ....... All other Properties come from some other Type!Wednesday, December 8, 2010
  • 21. /common/topic "Topics" • Things we have discourse about • Provides properties: • Alias • Article • Image • Weblinks • Assumed to be an "Included Type" for any "standard" typeWednesday, December 8, 2010
  • 22. Schema Patterns Compound Value Mediator Phylogeny EnumerationWednesday, December 8, 2010
  • 23. Compound Value complex literal Two or more properties which can only be interpreted with regard to one another Population • Dated Integer ("when did this location have that many people") Movie Budget • Dated money value • Date, Currency, Amount Ticker Symbol • Exchange, SymbolWednesday, December 8, 2010
  • 24. Compound Value currency estimated_budget amount 31MM {   "id":   "/en/apocalypse_now", valid_date 1979   "type": "/film/film",   "estimated_budget": [{     "currency":   null,     "amount":     null,     "valid_date": null   }] }Wednesday, December 8, 2010
  • 25. Mediator combine date annotation and character = tv character An annotation on the link between two Topics • Requires an object between the two Topics • The Topics become separated by two properties actor performance film character • Also useful for indicating the dates when a relationship existed (e.g., education, employment, etc.)Wednesday, December 8, 2010
  • 26. Mediator {   "id":   "/en/marie_antoinette_2006",   "type": "/film/film",   "starring": [{     "actor":null,     "character":null   }] }Wednesday, December 8, 2010
  • 27. Phylogeny Used when instances form a hierarchy Phylogeny properties have an expected Type which is the same as the Type on which the property is defined. Examples: • /location/location/containedby • /computer/computer/parent_model • /tv/tv_program/spin_offsWednesday, December 8, 2010
  • 28. Phylogeny {   "id": "/en/fairfax_california",   "/location/location/containedby": [{     "id": null,     "containedby": [{       "id": null Why can I use the short name??     }]   }] }Wednesday, December 8, 2010
  • 29. Enumerated Value Closed collection of “values” for a property Constrains relations to fixed set of objects • /people/person/gender { female, male, other } • /visual_art/visual_artist/art_forms { drawing, painting, print making, photography.... }Wednesday, December 8, 2010
  • 30. Explore the Freebase Graph child wrote directed parent sibling directed starred_in directed starred_inWednesday, December 8, 2010
  • 31. Explore the Freebase Graph [{   "id":   null,   "type": "/film/director" }]Wednesday, December 8, 2010
  • 32. Explore the Freebase Graph [{   "id":   null,   "type": "/film/director",   "/people/person/children": [{      "id":   null,      "type": "/film/director"   }] }]Wednesday, December 8, 2010
  • 33. Explore the Freebase Graph [{   "id":   null,   "type": "/film/director",   "film":[ ],   "/people/person/children": [{      "id":   null,      "type": "/film/director"      "film":[ ]   }] }]Wednesday, December 8, 2010
  • 34. Explore the Freebase Graph [{   "id":   null,   "type": "/film/director",   "film": [ ],   "/people/person/children": [{      "id":   null,      "type": "/film/director",      "film": [{        "name":null,        "starring": [{          "actor": null        }]      }]   }] }]Wednesday, December 8, 2010
  • 35. namespace It’s all nodes / pe ople ( key y) ) (ke film domain type domain /people and links! type /film e insta anc nce type inst nam e y) (ke e typ "domain" son type film per (key name "property" ins ce type ) tan propert tan ies ins ce ins tan "type" ce instance type nam e type instance typ e property instance expected_type ce tan ins na ins m nce e tan insta ce "expected type" type property type type type typ type type e type properties na expected _type date_time me na na me me na m e "person" "date of birth" nce "film" instance instance insta "actor" e LEGEND film typ name "Blade Runner" /type/object starring /type/object + /common/topic film obj type film character performance /namespace e nam "Rick Deckard" outgoing incoming actor keyvalue (key) film actor out goin gp rop person e e rty typ type literal value nam th e _o f_bir date "Harrison Ford" 1942-07-13Wednesday, December 8, 2010
  • 36. Domains, Bases and Commons "commons" individuals "bases" promo te Blade Runner "domains"Wednesday, December 8, 2010
  • 37. Questions?! Docs: www.freebase.com/docs Wiki: wiki.freebase.com Mailing List: lists.freebase.comWednesday, December 8, 2010